Research

Papers and Research Output

BECA Dataset

Published in Scientific Data (2025): long-term cattle recognition dataset and baseline verification framework.

  • My contribution: participated in full data pipeline design, detection/pose/ReID evaluation protocol construction, and baseline modeling experiments.
  • Measurable output: co-built dataset subsets reported in the paper: 16,889 images covering 5,661 cattle and 12,172 labeled images for long-term tracking of 103 cattle.
  • Engineering value: established reusable evaluation workflow for long-cycle recognition tasks, including cross-dataset validation settings.

DOI: 10.1038/s41597-025-06326-5

CRODNet

Published in Computers and Electronics in Agriculture (2025): contactless top-view rotated cattle detection framework.

  • My contribution: built top-view cattle data processing and rotated object detection workflow, and implemented lightweight vision modeling modules.
  • Measurable output: paper-reported system-level results include about 70% parameter reduction and about 50% FLOPs reduction with at least 3% AP gain versus compared bottom-up approaches.
  • Engineering value: completed cascaded pipeline of detection, keypoint localization, and alignment for robust top-view scenarios.

DOI: 10.1016/j.compag.2025.111160

Narrative Complexity Analysis

Under review (2026): semantic-entropy-based computational analysis of narrative complexity in Xianxiao texts.

  • My contribution: designed semantic entropy modeling strategy, narrative complexity indicators, and full statistical evaluation process.
  • Measurable output: completed an end-to-end 4-stage pipeline (text preprocessing, feature extraction, entropy modeling, statistical testing) and full manuscript writing.
  • Engineering value: translated literary analysis problems into reproducible computational workflow for research-scale iteration.

Target journal: Journal of Computational Literary Studies

Software Achievements

Software Copyright Portfolio

Non-contact Beef Cattle Body Growth Tracking System Based on Machine Vision

2025.06 · Machine vision software copyright project

  • My contribution: implemented top-view cattle detection and growth tracking algorithm modules with complete inference workflow.
  • Optimization focus: improved detection-to-tracking handoff stability and structured long-cycle growth monitoring logic.
  • Deliverables: production-oriented algorithm module package and verifiable technical documentation for recruitment review.

EEG Hidden-frequency System Based on Microtexture and Closed-loop Control

2025.10 · EEG software copyright project

  • My contribution: implemented EEG microtexture feature extraction and closed-loop control algorithm modules.
  • Optimization focus: tuned preprocessing and feature-flow stability for repeated closed-loop triggering experiments.
  • Deliverables: reusable algorithm implementation and validation scripts for iterative BCI experimentation.
Resume

I position my portfolio to show how I create practical AI and software value.

Target roles: LLM Application Engineer Intern, AI Developer Intern, and AI Product Engineering Intern. This page summarizes my technical scope, project ownership, and measurable outcomes for fast hiring evaluation.

5 End-to-End Projects 2 Published Papers + 1 Under Review 2 Software Copyrights

What This Portfolio Demonstrates

I am deeply interested in software development, large language models, and applied AI systems. In project practice, I focus on turning concepts into usable products with reliable engineering decisions.

I have built strong programming fundamentals and problem-solving ability through implementation-heavy work. I pay close attention to code quality, maintainability, and system design, and I adapt quickly in collaborative environments.

  • Engineering Direction: AI product development with strong software discipline.
  • Execution Style: clear scope, measurable progress, and practical delivery.
  • Project Execution Evidence: independently completed multi-project engineering runs across Shapeville implementation, MiniMind full pipeline reproduction, and WeClone end-to-end deployment validation.
  • Long-Term Goal: grow on the LLM/AI track and build products with real-world value.